The script is for classification of a subset of MNIST handwritten digits database obtained from the url: http://pjreddie.com/projects/mnist-in-csv/
The procedure followed is as follows:
1.The training set was divided into training subset and Cross Validation set by random sampling.
2.Each set was then feature normalized.
3.A single hidden layer neural network using package "nnet" was trained on the training subset for various parameter values and corresponding CV set errors were stored.
4.The parameter value with least amount of CV set error was selected and was used to make predictions on the test set.
Please read "Project Summary.pdf" for detailed description.